How Machine Learning Can Make Medicine More Human

Picture of Sefton Eisenhart
Sefton Eisenhart

Artificial intelligence has long been synonymous with science-fiction. Some people see the capabilities of AI and are filled with hope at its potential to revolutionize fundamental elements of daily life. Other people regard AI with skepticism—even paranoia.

As with most spectrums of opinion, reality lies in the middle ground. No one can deny the fact that artificial intelligence will impact nearly every business in the world. Transportation, agriculture, finance and so many more, will be affected. Healthcare is no exception.

It’s impossible to quantify the amount of medical research in existence. Valuable insights are hidden in the vast amounts of information, and discovering the right information can change the way diseases are cured. AI is able to learn rules by analyzing huge amounts of repetitive data. Based on those rules, it finds insights. No human force can match the efficiency of AI when it comes to combing through material and finding points of interest.

While every disease is different, many share a large number of characteristics. Researchers often don’t know where to look to find commonalities—there is simply too much data and too little collaboration. AI can help sift through incredible amounts of past information, identify useful elements, and get them in front of stakeholders who can apply human expertise to a refined and realistic amount of information. It will never be enough to just run data through an AI application. There has to be an expert at the other end to make sense of the results and apply findings to research and development.

At TREND, we are trying to take AI applications even further, and use the technology to capture one of the most overlooked pieces in the research process—the patient voice. Natural history data is biased toward what doctors think is important about the disease. Conversational data, on the other hand, more likely concerns what actual patients and caregivers think is important.

Patient conversations were often dismissed as anecdotal data with little value in development process. But, when it comes to natural history or clinical data—especially electronic health records or scientific studies—it can be easy to lose the narratives of the people living with the disease. When researchers and stakeholders have complicated regulations and processes to contend with, they might not have the bandwidth to stay in touch with the communities they are working so hard to serve.

At TREND, we have been applying machine learning to the conversations of actual patients and caregivers. We collaborate with huge communities and find valuable insights in the things they say to each other. We find information you might not see in a rigid group of EHRs. We use AI to analyze tens of thousands of conversations, identify patterns, and make new discoveries based on what people are saying. We shine a light on nuances that go unnoticed in scientific studies. We amplify the voices of the people who actually feel the illness, and this data helps accelerate understanding on the road to a cure.